基于小波变换和BP神经网络的织物疵点检测  

Fabric Defect Detection Based on Wavelet Transform and BP Neural Network

作　　者：李鹏飞[1] 杜莹萍 张缓缓[1] 景军锋[1] 

机构地区：[1]西安工程大学,陕西西安710048

出　　处：《棉纺织技术》2014年第10期21-26,共6页Cotton Textile Technology

基　　金：国家自然科学基金项目(61301276);陕西省教育厅专项(2013JK1084)

摘　　要：探讨基于小波变换和BP神经网络的织物疵点检测技术。为准确检测织物疵点,采用小波变换对预处理后的织物图像进行分解,小波分解后不同的子图像反应了织物的不同细节信息,从小波分解后的水平细节子图像和垂直细节子图像中提取特征参数,特征参数的提取采用灰度共生矩阵法,将提取到的特征参数送入训练过的BP神经网络,进行检测疵点,达到疵点织物融合、形态学和阈值处理并显示疵点的目标。实验证明:该方法行之有效。认为:寻找更适合的方法提取更有效的特征值和改进神经网络可以提高识别效率。Fabric defect detection technology based on wavelet transform and BP neural network were discussed. To detect fabric defect accurately, fabric image pretreated by wavelet transform was decomposed. Different detail informa- tion of the fabric can be reflected by decomposed sub-images. Characteristic parameters were extracted from horizontal detail sub-images and vertical detail sub-images. Gray-level co-occurrence matrix was used to extract characteristic pa- rameters. The characteristic parameters were input trained BP neural network to detect fabric defect, aims of mixing de- fect and fabric, dealing of morphology and threshold, displaying of defects can be reached. The test shows that the method is effective. It is considered that fabric defect recognition efficiency can be increased through finding better method to extract more effective characteristic parameters and modify BP neural network.

关 键 词：小波变换 BP神经网络 灰度共生矩阵 疵点检测 特征值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]